Air-quality-prediction

Overview

• Prediction of pollution indicators - PM2.5, PM10, SO2 and NO2
• Predicting 1, 2 and 3 days into the future using past 24 hours data only
• Trained using keras with tensorflow as backend

Dependencies required

• scikit-learn
• keras tensorflow
• numpy, pandas, matplotlib, seaborn and tqdm

Steps for training model

• open script.ipynb and follow instructions
• change hyperparamters in trainmodel functions

Steps for using pretrained Model

• functions provided in demo.py for preprocessing and predicting using trained model and printing metrics and ploting graphs
• run_CNN(pollutant='PM2.5',days=1)
• run_ANN(pollutant='PM2.5',days=1)
• run_SVM(pollutant='PM2.5',days=1)
• run_LR(pollutant='PM2.5',days=1)
• pollutant strings parameters from 'PM2.5', 'PM10', 'NO2', 'SO2'
• days argument can be 1, 2 or 3